Gabor Szabo | 5 Jun 2005 22:20
Picon
Gravatar

Conflict on deleted file

svk push sais I have 3 confilcts on a directory and two files in it
but I actually already deleted those files and even the directory using svk del.

So now 

svk resolved . -R 
svk resolved dirname -R 

do not seem to do anything.

So how do I resolve the conflict now ?
How can I use svk push again ?

Gabor

Gabor Szabo | 6 Jun 2005 00:36
Picon
Gravatar

Re: Conflict on deleted file

after some advide on IRC #svk I was still at odds with this problem
I tried the same commands in several order but nothing seemed to work.
the delete files were gone from the workspace and the local branch but were
still available in the remote repository and in the local mirror of
the remote repository.
copied back the directory:

svk copy //remote/somedir  //local/somedir

and the deleted the files using
      svk del filename
again and ran svk smerge //local/somedir //remote/somedir

This elinated the conflict on the two files (got lower case d-s when tried to
push the directoy) but now I had two conflicst on
somedir   and on
somedir/

(yes, I had a slash difference)

In the end I ran
    svk del //remote/somedir
after that push-ing already worked and it seems everything was finally updated
in the remote repository.

Gabor

solo turn | 8 Jun 2005 10:44
Picon

Re: Re: password problem with svn 1.2

On 6/7/05, John Peacock <jpeacock <at> rowman.com> wrote:
> solo turn wrote:
> > ssl does not work, without ssl it works. i did not yet try an older
> > svn (cause it used to work before ...).
> 
> Are you sure you built svn with SSL support:
> 
> $ svn --version
> svn, version 1.1.4 (r13838)
>     compiled Apr 14 2005, 17:04:54
> 
> ...snip...

with svn it works (we just took the jostein build from tigris), with
svk it does not any more, and we did not change the server and the
access rights. if ssl is not supported the error is not "authorization
required".

the strange thing is, that with http it works, and with https we get
we get a "wrong password" error on the server.

i will try an older svn version, and also a different server version
(was and still is 1.2.0-dev unchanged for a few months).

-solo.

solo turn | 11 Jun 2005 21:13
Picon

codeville merge

hi,

what are your plans with merging in svk? are they somehow similar to
codeville and monotone (like stated in
http://article.gmane.org/gmane.comp.version-control.monotone.devel/3264),
or does svn repository disallow such an approach?

-solo.

Marijn Vriens | 12 Jun 2005 10:16
Picon

svk and feature-branches

Hi all,

Great stuff, SVK, however I have a question..

My situation: I want to collaborate with a project that uses a normal
SVN repository to manage their code. I don't have rights to commit to
trunk/. But, I do have my own directory (/users/marijn/) in that
repository where I can commit to.

I have various tasks to do, some short running, and some that will take
me a couple of weeks to finish. 

My Idea is to take advantage of the cheap copies of SVN and have a copy
of trunk/ (a branch) for each of the tasks. So that the person merging
back into trunk/ can pick the features they want to have in.

Since some of the tasks will be longer running, it would be really nice
to be able to follow (merge) the changes happening in trunk, while still
having my changes in my own branch. So that once I am finished, i can
say: "here's a branch that is trunk <at> HEAD plus this extra
feature" (making life that much easier on the integrator)

So how I trying to do this right now is:
- I mirror the whole of the SVN repository into //mirror/project (this
contains both trunk/ and users/marijn/)
- Then I copy //mirror/project to //local/project
- copy //mirror/project/trunk/ to //local/project/users/marijn/task
- Checkout //local/project/users/marijn/task
- Get to work

(Continue reading)

David Glasser | 13 Jun 2005 05:18
Favicon

PATCH: svk co --purge, and indicate nonexistentant in svk co --list

Attachment (svk-checkout-purge.patch): application/octet-stream, 11 KiB
The attached path adds an "svk checkout --purge" command which  
detaches checkout directories that no longer exist (after prompting),  
and adds an indicator to the listing in "svk checkout --list" showing  
whether or not the directory exists.

Please note that while all of the test results of "svk checkout -- 
list" that I've added are internally consistent, I am not sure that  
the first one is actually correct.  Somebody with a little more svk  
experience than me should take a look at the first --list that I  
added (grep for 'sanity-checked'), and make sure that all of the  
entries are actually correct!

These features were originally implemented by knewt; I mostly just  
updated his patch to work with the latest svk and added the prompting  
and tests.

--dave
   Code Monkey, Best Practical Solutions
--

-- 
David Glasser | glasser <at> bestpractical.com

Rahul Jain | 14 Jun 2005 09:36
Picon

Replication and Synchronization of Subversion repositories

Hi:

We are using subversion as our source code control system. We are
opening a new office in another country shortly where we will work on
the same codebase as the current office. We want to have a separate
subversion repository server in the new office with the same codebase
as the current office.

We are thinking that the employees in the old office should keep using
the repository server in their office and the employees in the new
office will use the server in the new office. There should be a
synchronization script that will synchronize the data between the two
repositories at regular intervals.

Subversion does not provide this feature out of the box. Can we use
SVK to achieve this? Has someone used this in a production-worthy
environment? Any pointers on how to achieve this would be very helpful
and appreciated.

Thanks,
Rahul.

Martin Blom | 14 Jun 2005 23:08

Re: SVN::Mirror::Ra::MirrorEditor error?

Chia-liang Kao wrote:
> On Fri, May 06, 2005 at 07:03:30PM +0200, Martin Blom wrote:
> 
>>Chia-liang Kao wrote:
>>
>>>>>On Wed, May 04, 2005 at 05:52:45PM +0200, Martin Blom wrote:
>>>
>>>
>>>
>>>With some ktrace help, I find the libsvn_subr's io_remove_dir is missing
>>>one entry, so the rmdir(2) call failed. And this looks very familiar
>>>to me.
>>>
>>>can you try recompile svn with the MACOSX_REWINDDIR_HACK thing
>>>found in subversion/libsvn_subr/io.c and see if that helps?
>>
>>Sorry, same behaviour as before. It stops/crashes on
>>
> 
> 
> Can you send me strace output on syscalls?

Sorry for my late reply; I've been away from coding for a while. I've 
retried with SVK 1.00 as shown below. As before, at revision 284 it just 
stops (without any error messages at all) after which the repository is 
messed up.

[martin <at> lisa ~]$ rm -rf .svk [martin <at> lisa ~]$ svk depotmap --init
Repository /home/martin/.svk/local does not exist, create? (y/n)y
[martin <at> lisa ~]$ svk mirror svn+ssh://svn.berlios.de/svnroot/repos/arp2 
//mirror/arp2
Committed revision 1.
[martin <at> lisa ~]$ strace -o /tmp/svk-mirror-2.log svk sync //mirror/arp2
Syncing svn+ssh://svn.berlios.de/svnroot/repos/arp2
Retrieving log information from 1 to 658
Committed revision 2 from revision 1.
[...]
Committed revision 284 from revision 290.
[martin <at> lisa ~]$ strace -o /tmp/svk-mirror-3.log svk sync //mirror/arp2
Waiting for mirror lock on //mirror/arp2: lisa.datan.blom.org:16126.
Waiting for mirror lock on //mirror/arp2: lisa.datan.blom.org:16126.
Waiting for mirror lock on //mirror/arp2: lisa.datan.blom.org:16126.
[...]
[martin <at> lisa ~]$ svn --version
svn, version 1.1.4 (r13838)
    compiled May  6 2005, 18:28:50

Copyright (C) 2000-2004 CollabNet.
Subversion is open source software, see http://subversion.tigris.org/
This product includes software developed by CollabNet 
(http://www.Collab.Net/).

The following repository access (RA) modules are available:

* ra_dav : Module for accessing a repository via WebDAV (DeltaV) protocol.
   - handles 'http' schema
   - handles 'https' schema
* ra_local : Module for accessing a repository on local disk.
   - handles 'file' schema
* ra_svn : Module for accessing a repository using the svn network protocol.
   - handles 'svn' schema

[martin <at> lisa ~]$ svk -v
This is svk, version 1.00.
[martin <at> lisa ~]$

I have uploaded the strace logs to http://martin.blom.org/tmp/svk/ 
(about 1.5 MB compressed).

--

-- 
---- Martin Blom --------------------------- martin <at> blom.org ----
Eccl 1:18                                 http://martin.blom.org/

Martin Blom | 14 Jun 2005 23:14

Re: SVN::Mirror::Ra::MirrorEditor error?

Stefan Kuttler wrote:

>>>Meanwhile can you check if this only fails on nfs?
>>
>>Yes, this is definately related to NFS. I have no such problems when I 
>>run SVK on the server itself.
> 
>  
> Maybe internal locking conflicts with NFS locks?
> Just a thouhgt, might be comletely, uhm, wrong :-)

It looks like it fails because there is a .nfsXXXXX file left. Why, I 
don't know, but it usually means that a file was deleted while still 
being open by another process.

--

-- 
---- Martin Blom --------------------------- martin <at> blom.org ----
Eccl 1:18                                 http://martin.blom.org/

sasha.ov | 15 Jun 2005 07:22
Picon

Re: Replication and Synchronization of Subversion repositories

Rahul --

I am in a similar situation, with 2 geographically distributed repositories.

I am afraid there may be no automatic way to sycnrhonize 2 repositories on the same branch of development. When all the engineers work against the single repository, the conflicts are resolved at the time of check-in. When repositories are 2 or more, the conflicts need to be resolved at the time of synchronization, so you will somehow need a human in the middle of this process.

The options I am considering for my situation:
  • Each engineer in the smaller office uses SVK to mirror/mergeback changes against the central repo. The repo is located in the bigger office, where engineers work locally. In my experience, SVK performs beautifully on slow connections even if the repository is big, so performance shouldn't be a bottleneck here.
  • Each office checks into its own repository and there is 1 designated person to oversee synchronization (under the condition this person would not be me ;-)
  • Each office checks into its own branch. As there may be no conflicts, the branches are synchronized automatically (possibly using SVK). From time to time somebody in each office does incremental merge between the branches (using SVK -- a piece of cake) and checks modification into their branch.
There is no clear cut, each option has its advantages/disadvantages, so I am still looking. Will be happy to learn about other possibilities.

Regards,
Sasha


Rahul Jain wrote on 06/14/2005 12:36 AM:
Hi: We are using subversion as our source code control system. We are opening a new office in another country shortly where we will work on the same codebase as the current office. We want to have a separate subversion repository server in the new office with the same codebase as the current office. We are thinking that the employees in the old office should keep using the repository server in their office and the employees in the new office will use the server in the new office. There should be a synchronization script that will synchronize the data between the two repositories at regular intervals. Subversion does not provide this feature out of the box. Can we use SVK to achieve this? Has someone used this in a production-worthy environment? Any pointers on how to achieve this would be very helpful and appreciated. Thanks, Rahul.

Gmane